Elon Musk’s xAI Releases Grok 3 in API for Developers, Introduces a ‘Fast’ Model

Elon Musk's xAI, the artificial intelligence (AI) firm behind the Grok chatbot, released a new application programming interface (API) on Thursday. The new API is for the Grok 3 AI model, which the company introduced in February. So far, the AI model has been available to the end consumer via the X (formerly known as Twitter) platform, the dedicated website, or the mobile apps. However, developers did not have access to the latest models. With this, developers will be able to build applications and software using the Grok 3 AI model.

xAI Releases Grok 3 API, Reveals Pricing

On its documents page, xAI detailed the new API, which will be powered by the Grok 3 AI models. The page contains model overviews, pricing details, and additional information that will be important for developers who access these models. In total, there are four large language models (LLMs) that developers will be able to access via the new API — Grok 3 and Grok 3 mini (both available in beta), as well as two faster variants of the same models.

The Grok 3 model is the company's flagship AI model that is also powering the chatbot on X, as well as the apps. xAI said that this model can be helpful in enterprise use cases such as data extraction, coding, and text summarisation. On the other hand, the Grok 3 mini is a lightweight model with native reasoning capability. The AI firm claims it is suited for logic-based tasks that “do not require deep domain knowledge.”

Additionally, Grok 3 Fast beta and Grok 3 mini Fast beta are also available with the API. xAI said both of these use the same underlying model as the non-fast variant, and deliver identical response quality. The main difference is that the faster variants are served on special, faster infrastructure and offer significantly faster response times compared to the standard model. As a result, these models also have a higher input and output pricing.

Coming to pricing details, the Grok 3 model will cost developers $3 (roughly Rs. 260) per million input tokens and $15 (roughly Rs. 1,300) per million output tokens. On the other hand, the faster variant comes at the price of $5 (roughly Rs. 430) per million input and $25 (roughly Rs. 2,150) per million output tokens.

The mini models are considerably cheaper compared to their standard counterparts. Grok 3 mini is priced at $0.30 (roughly Rs. 25) per million input tokens, and $0.50 (roughly Rs. 43) per million output tokens. The Grok 3 mini fast AI model will cost developers $0.60 (roughly Rs. 51) for every million input tokens, and $4 (roughly Rs. 344) for every million output tokens.

Notably, these AI models are text-only and cannot generate images. However, the models do accept images as input. Additionally, the API version of Grok 3 is not connected to the Internet, and the models are reliant on their pre-training datasets.